EMC Annual Review:CPC’s Forecasts
2007
Edward O’LenicChief, Operations Branch
NOAA-NWS-Climate Prediction CenterDecember 12, 2007
Top 20 CPC Web Products – Nov 20071.) CPC Home page (150,730 hits) 2.) 6-10 Day Outlook (126,958) 3.) 8-14 Day Outlook (119,752) 4.) U.S. Hazards Assessment (58,890) 5.) ENSO Evolution, Status.....(PDF) (52,864) 6.) 90-Day Outlook (46,558) 7.) Forecasts & Outlooks Home page (45,625) 8.) 6-10 Day Outlook (PMD) (38,491) 9.) 30-Day Outlook (32,109)
10.) UV Index Forecast (text version) (31,487) 11.) ENSO Diagnostic Discussion (31,011) 12.) 30-Day Outlook (Lead 01) (29,214) 13.) Color Seasonal Outlook (multiple images) (27,224) 14.) Seasonal Drought Outlook (26,120) 15.) U.S. Daily Data (text) (20,957) 16.) 30-Day Outlook (Lead 02) (18,993) 17.) ENSO Home page (18,506) 18.) Seasonal Drought Assessment (14,736) 19.) Atlantic Hurricane Outlook (13,030) 20.) 90-Day Outlook (fxus05, text) (13,005)
Top 20 CPC Web Products – Nov 20071.) CPC Home page (150,730 hits)2.) 6-10 Day Outlook (126,958) 3.) 8-14 Day Outlook (119,752) 4.) U.S. Hazards Assessment (58,890) 5.) ENSO Evolution, Status.....(PDF) (52,864)6.) 90-Day Outlook (46,558) 7.) Forecasts & Outlooks Home page (45,625) 8.) 6-10 Day Outlook (PMD) (38,491) 9.) 30-Day Outlook (32,109)
10.) UV Index Forecast (text version) (31,487) 11.) ENSO Diagnostic Discussion (31,011) 12.) 30-Day Outlook (Lead 01) (29,214) 13.) Color Seasonal Outlook (multiple images) (27,224)14.) Seasonal Drought Outlook (26,120) 15.) U.S. Daily Data (text) (20,957) 16.) 30-Day Outlook (Lead 02) (18,993) 17.) ENSO Home page (18,506) 18.) Seasonal Drought Assessment (14,736) 19.) Atlantic Hurricane Outlook (13,030) 20.) 90-Day Outlook (fxus05, text) (13,005)
Outline• Atlantic Tropical Storm Outlook• U.S. Seasonal Drought Outlook• U.S. Hazards Assessment• Extended Range – 6-10, 8-14-Day • Intra-Seasonal Outlooks• 3-Month Outlooks
2007 Atlantic Tropical Storm Season
May 16 Forecast
August 2Forecast
2007 Actual
AverageSeason
TropicalStorms
11
62
Hurricanes
MajorHurricanes
Obs.% Normal
13-17 13-16 14 -27
7-10 7-9 6 1003-5 3-5 2 100
2007 was close to average. After several successful forecasts, this is the second consecutive year in which the forecast’s
expectations were not met.
U.S. Seasonal Drought Outlook• DM is produced collaboratively each by
NOAA, USDA, NDMC.• Seasonal Drought Outlook is produced
monthly at CPC.• Outlook is zero-lead.• Describes expected changes to the DM,
based on CPC’s seasonal outlook.• Uses up-to-the-minute model forecasts, as
well as the 3-month P outlook.
% DO gridpoints correct minus % persistence gridpoints correct
Drought Outlook -- Forecast minus Persistence
-20
-10
0
10
20
30
40
50
60
Jul '03 Nov'03
Mar'04
Jul '04 Nov'04
Mar'05
Jul '05 Nov'05
Mar'06
Jul '06 Nov'06
Mar'07 Jul'07
Date of Forecast
Fore
cast
Les
s Pe
rsis
tenc
e
Skill MEAN Linear (Skill)
Average: 13 % better than climatology
U.S. Hazards Assessment
Observed Not observed
Forecast 2086
a
7306b
Not forecast 41753
c307685
d
CONTINGENCY TABLEHEAVY PRECIPITATION HAZARDS
2002-2007
Hit Rate, heavyPrecipitation events
= a/(a+c)=# hits/(# fcst and obs)= 0.05
False Alarm rate, heavyPrecipitation events
= b/(b+d)=# miss/(# all forecasts)= 0.02
Heavy precipitation is defined as the greater of the 95th percentile, or 1 inch per day.
Updated each Mon-Fri, using daily GFS forecasts of extreme events.
Bias, heavyPrecipitation events
= (a+b)(a+c)=# yes fcsts/(# yes obs)= 0.21
U.S. Hazards Assessment
Observed Not observed
Forecast 2086
a
7306b
Not forecast 41753
c307685
d
CONTINGENCY TABLEHEAVY PRECIPITATION HAZARDS
2002-2007
Heavy precipitation is defined as the greater of the 95th percentile, or 1 inch per day.
Updated each Mon-Fri, using daily GFS forecasts of extreme events.Needed: Calibrated extreme event forecast probabilities from the GEFS
Hit Rate, heavyPrecipitation events
= a/(a+c)=# hits/(# fcst and obs)= 0.05
False Alarm rate, heavyPrecipitation events
= b/(b+d)=# miss/(# all forecasts)= 0.02
Bias, heavyPrecipitation events
= (a+b)(a+c)=# yes fcsts/(# yes obs)= 0.21
NCEP- CPC Extended-Range (6-10-, 8-14-Days) Forecast Operations
OCEAN
ANALYSIS, RE-ANALYSIS
DYNAMICAL Models:
NCEP GFS, others… STATISTICAL
T, P TOOLS, CALIBRATION
OFFICIAL TEMPERATURE & PRECIPITATION
OUTLOOKS
Forecasters merge T, P tools subjectively
ATMOSPHERE
ANALYSIS, RE-ANALYSIS
Atmosphere forecasts
VERIFICATION
Objective 500-hPa Height forecast, from forecaster-
assigned weights
OFFICIAL 500 HPA HEIGHT AND
ANOMALY FORECAST
VERIFICATION
8-14-day (D+11) Forecast for Dec 19-25
ER Z-500 ACOfficial, GFS
Height forecasts are made by calculating a weighted average of GFS, ECMWF, Canadian models, with weights selected by the forecaster. GFS, Canadian, ECMWF are the choices. GFS is usually the major component. No subjective modifications are made to the weighted mean by the forecaster.
6-10
8-14
Calibration (Downscaling) is Extremely Important in Climate ForeCalibration (Downscaling) is Extremely Important in Climate Forecastingcasting
Ensemble Regression Assembles an Optimal Distribution Based on Spread of the Members and on
the Skill of the Forecast system.
Official ER Forecast Heidke SS, 6-10-/8-14-Day T, PMonthly Means, Dec 2006-Nov 2007
-20
-10
0
10
20
30
40
50
60
200612 200701 200702 200703 200704 200705 200706 200707 200708 200709 200710 200711
Month
% Im
prov
emen
t Ove
r Ran
dom
610 T
814 T
610 P
814 P
mean=38.2
mean=29.2
mean=15.5
mean=4.6
DEC2006
NOV2007
Daily 6-10- and 8-14-day Heat Index (Wind Chill) Outlooks are prepared by calculating the heatindex (wind chill) from GFS 2-meter T, RH (GFS 2-meter T, 10-meter wind) and then calibrating
using 45 days of observed heat index (wind chill) and GFS forecasts.
Probability theminimum wind chill will be in the below -10F.
May Oct
Oct Apr
WARM SEASON
COLD SEASON
NCEP- CPC Intra-Seasonal Forecast Operations
OCEAN
ANALYSIS, RE-ANALYSIS
DYNAMICAL Models:
NCEP GEFSSTATISTICAL T, P TOOLS
EXPERIMENTAL GLOBAL HAZARDS
MAP AND TEXT
Forecasters merge T, P tools subjectively
ATMOSPHERE
ANALYSIS, RE-ANALYSIS
Atmosphere forecasts
VERIFICATION
GEFS projected onto MJO CEOF modes
VERIFICATION
Experimental MJO ForecastsDaily GEFS
forecasts are projected onto modes resulting from a CEOF analysis of U200, U850, OLR.
GEFS Based MJO Forecast Tool
Current GEFS MJO index forecast
Recent MJO Event Verification
Used as part of weekly CPC MJO weekly update and hazard assessment products. Recently, GEFS MJO forecasts have been remarkably accurate. Forecast (green) begun December 3,
2007
Forecast (green) begun Nov 26, 2007. Others are observations.
GFSCFS
Verification, January-September, 2007The success of the GEFS in forecasting the latest MJO
reflects, in part, the impact of the information in the initialization. Ensemble Regression can be applied to the ensemble members.
Global Tropics Hazards/Benefits AssessmentWeekly MJO Update
NCEP- CPC Seasonal Forecast Operations
OCEAN
ANALYSIS, RE-ANALYSIS
DYNAMICAL MODEL TOOLS:
NCEP CFS, OTHERS …
STATISTICAL SST MODEL
TOOLS
STATISTICAL T, P MODEL
TOOLS
OBJECTIVE CONSOLIDATION BASED ON KNOWN SKILL OF FORECAST TOOLS (ensemble
regression)
SUBJECTIVE FORMULATION OF FORECAST MAPS
OBJECTIVE CONSOLIDATION BASED ON KNOWN SKILL OF TOOLS (ensemble regression)
ATMOSPHERE
ANALYSIS, RE-ANALYSIS
OFFICIAL SST FCST
Atmosphere
forecastsO
cean
fore
cast
s
OFFICIAL TEMPERATURE & PRECIPITATION OUTLOOKS
VERIFICATION
VERIFICATION
CRPS Skill
00.10.20.30.40.50.6
1 2 3 4 5 6
Lead (Months)
CR
PSS
CCA CA MKV CFS-Uncal CFS ALL
1995-2006Ensemble Regression Calibration Improves CFS SST Forecasts
CFS Nino 3.4 SST Forecasts 1995-2006
CFS & Official SST Forecasts
Heidke Skill Score 48 Month Running Mean Official 3-month T OutlooksNDJ 1998-SON 2007
0.00
5.00
10.00
15.00
20.00
25.00
30.00
35.00
Dec-98
Mar-99
Jun-99
Sep-99Dec-9
9Mar-0
0Ju
n-00Sep-00Dec-0
0Mar-0
1Ju
n-01Sep-01Dec-0
1Mar-0
2Ju
n-02Sep-02Dec-0
2Mar-0
3Ju
n-03Sep-03Dec-0
3Mar-0
4Ju
n-04Sep-04Dec-0
4Mar-0
5Ju
n-05Sep-05Dec-0
5Mar-0
6Ju
n-06Sep-06Dec-0
6Mar-0
7Ju
n-07Sep-07
Ending Month of 3-month Forecast
% Im
prov
emen
t ove
r ran
dom
GPRA GPRA GOAL
SON 2005:First forecast made using
the consolidation
Regional % Improvement by the ½-month Lead 3-Month Mean Temperature Forecast over Climatology
Official and Consolidation – 1995-2004
Difference: Official minus Consolidation
Prospects for the (Present &) Future
• Probabilistic Extreme Events forecasts (U.S. Hazards Assessment)• Disaggregated ½-Month Lead 3-Month Temperature Outlooks• Operational MJO Outlooks for the Tropics• CPC Forecasts and Climatologies on NDFD available to users via SOAP• Gridded 3-Month Outlooks for the Pacific and Caribbean• Expansion of Ensemble Regression Calibration to Alaska 3-Month Outlooks• Probability of Exceedance for ER forecasts